Fix the month detection for lowercase and long abbreviated month names. Fix ddd and mmm formats to make them fit more locales.

[Exoth]

Let's try the usual case:

>> I18n.locale = :en
:en

>> "18 October 2013".to_date
Fri, 18 Oct 2013

>> "18 october 2013".to_date
Fri, 18 Oct 2013

>> "18 oct 2013".to_date
Fri, 18 Oct 2013

>> "18 Oct 2013".to_date
Fri, 18 Oct 2013

So far everything is ok. But in fact there are a lot of locales, which have lowercase and long abbreviated month names, like 'octobre' and 'oct.' in French.

>> I18n.locale = :fr
:fr

>> "18 Octobre 2013".to_date
ArgumentError: invalid date
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/bundler/gems/timeliness-9c8161f4ab08/lib/timeliness/core_ext/string.rb:12:in `new'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/bundler/gems/timeliness-9c8161f4ab08/lib/timeliness/core_ext/string.rb:12:in `to_date'
  from (irb):39
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:47:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:8:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands.rb:41:in `<top (required)>'
  from /home/exoth/projects/jetradar/script/rails:6:in `require'
  from /home/exoth/projects/jetradar/script/rails:6:in `<top (required)>'
  from -e:1:in `load'
  from -e:1:in `<main>'

>> "18 octobre 2013".to_date
ArgumentError: invalid date
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/bundler/gems/timeliness-9c8161f4ab08/lib/timeliness/core_ext/string.rb:12:in `new'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/bundler/gems/timeliness-9c8161f4ab08/lib/timeliness/core_ext/string.rb:12:in `to_date'
  from (irb):41
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:47:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:8:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands.rb:41:in `<top (required)>'
  from /home/exoth/projects/jetradar/script/rails:6:in `require'
  from /home/exoth/projects/jetradar/script/rails:6:in `<top (required)>'
  from -e:1:in `load'
  from -e:1:in `<main>'

>> "18 oct. 2013".to_date
NoMethodError: undefined method `map' for nil:NilClass
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/bundler/gems/timeliness-9c8161f4ab08/lib/timeliness/core_ext/string.rb:11:in `to_date'
  from (irb):43
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:47:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:8:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands.rb:41:in `<top (required)>'
  from /home/exoth/projects/jetradar/script/rails:6:in `require'
  from /home/exoth/projects/jetradar/script/rails:6:in `<top (required)>'
  from -e:1:in `load'
  from -e:1:in `<main>'

>> "18 Oct. 2013".to_date
NoMethodError: undefined method `map' for nil:NilClass
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/bundler/gems/timeliness-9c8161f4ab08/lib/timeliness/core_ext/string.rb:11:in `to_date'
  from (irb):45
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:47:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands/console.rb:8:in `start'
  from /home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/railties-3.2.14/lib/rails/commands.rb:41:in `<top (required)>'
  from /home/exoth/projects/jetradar/script/rails:6:in `require'
  from /home/exoth/projects/jetradar/script/rails:6:in `<top (required)>'
  from -e:1:in `load'
  from -e:1:in `<main>'

So that pull request fixes the month detection behavior to properly handle such situations.

And the second part is about regexp used for ddd and mmm formats.

Here's a simple code, checking how well the regexps perform against translations from all the locales found in the rails-i18n gem:

def check key, regexp
  bad = []
  locales = Dir['/home/exoth/.rvm/gems/ruby-1.9.3-p448@jetradar/gems/rails-i18n-0.7.4/rails/locale/*']
  locales.each do |file| 
    q = YAML.load(File.read(file))
    bad << q.keys.first if q.values.first["date"][key].compact.any?{|i| i !~ regexp} 
  end
  puts "#{bad.count}/#{locales.count} miss the regexp: #{bad.join(', ')}."
end

>> check 'month_names', /^\w{3,9}$/
check 'month_names', /^\w{3,9}$/
59/80 miss the regexp: es-AR, th, es-VE, el, is, az, de, tr, he, cy, zh-CN, fr-CH, lv, ca, hi-IN, pl, et, zh-HK, fr, mk, hi, hr, sk, uk, lt, es-PE, ro, pt, fa, pt-BR, lo, ne, es-419, kn, gl, ar, hu, fi, cs, es, sw, zh-TW, ja, es-MX, es-CL, eo, es-CO, bg, sr, or, de-AT, ru, wo, fr-CA, ko, vi, bn, de-CH, mn.

>> check 'abbr_month_names', /^\w{3,9}$/
43/80 miss the regexp: th, el, is, az, de, tr, he, zh-CN, fr-CH, lv, ca, uz, pl, et, zh-HK, fr, mk, hi, hr, sk, uk, fa, lo, ne, gl, ar, hu, fi, cs, zh-TW, ja, eo, bg, sr, or, de-AT, ru, fr-CA, ko, vi, bn, de-CH, mn.

>> check 'day_names', /^\w{3,9}$/
66/80 miss the regexp: es-AR, th, es-VE, el, sl, is, az, de, tr, he, cy, sv, zh-CN, id, lv, nb, uz, hi-IN, pl, et, zh-HK, mk, hi, hr, sk, uk, lt, es-PE, it, ro, nn, bs, pt, it-CH, fa, pt-BR, da, lo, ne, tl, es-419, kn, gl, ar, hu, fi, eu, cs, es, zh-TW, ja, es-MX, es-CL, eo, es-CO, bg, sr, or, de-AT, ru, wo, ko, vi, bn, de-CH, mn.

>> check 'abbr_day_names', /^\w{3,9}$/
check 'abbr_day_names', /^\w{3,9}$/
61/80 miss the regexp: es-AR, th, es-VE, el, sl, is, az, de, tr, he, sv, zh-CN, rm, lv, ca, nb, uz, hi-IN, pl, et, zh-HK, mk, hi, hr, sk, uk, lt, es-PE, ro, nn, bs, pt, fa, pt-BR, da, lo, ne, es-419, kn, ar, hu, fi, cs, es, sw, zh-TW, ja, es-MX, es-CL, eo, es-CO, bg, sr, or, de-AT, ru, ko, vi, bn, de-CH, mn.

So there are quite a lot of misses. So far these are the best results I've managed to get by making minor changes to the regexps:

>> check 'month_names', /^[[[:alnum:]]\.]{2,11}$/
11/80 miss the regexp: th, hi-IN, hi, ne, kn, sw, or, wo, vi, bn, mn.

>> check 'abbr_month_names', /^[[[:alnum:]]\.]{2,11}$/
8/80 miss the regexp: ca, hi, lo, ne, or, vi, bn, mn.

>> check 'day_names', /^[[[:alnum:]]\.]{1,14}$/
13/80 miss the regexp: th, az, cy, id, hi-IN, hi, uk, fa, ne, kn, or, vi, bn.

>> check 'abbr_day_names', /^[[[:alnum:]]\.]{1,14}$/
8/80 miss the regexp: hi-IN, hi, lo, ne, kn, or, vi, bn.

Much better already, though there are still some locales left, but they mostly involve multiword names, and this probably can't be fixed just by changing the format regexps. The way I see is to use the list of month/day names of the current locale at the stage of format processing.

Anyway, the pull request includes the changes in formats as specified above.

[Exoth]

Now when this solution couldn't match the Thai February month name translation I decided to go even further and defined mmm like this:

months = I18n.available_locales.flat_map do |locale| 
  I18n.t('date.month_names', locale: locale) + I18n.t('date.abbr_month_names', locale: locale)
end 
months.compact.uniq.map { |month| Regexp.escape(month) }.join('|')

and enabled case insensitive search. This way I'm guaranteed to match any possible month translation. But regexp is much bigger now and I don't know how it affects the performance. Anyway tell me if you find this solution suitable for the timeliness gem and I'll make a pull request.